基于遗传算法的决策树组合方法及在CRM的应用

针对客户关系管理中的客户分类问题,提出一种基于遗传算法的多重决策树组合分类方法.组合分类方法将概率度量水平的多重决策树并行组合,组合算法中采用遗传算法优化连接权值矩阵,并且采用客户信用等级评定的仿真数据对该方法进行测试和评估.实验结果表明,该组合分类方法比单个决策树及其他组合方法具有更高的分类精度,并在保持分类结果良好可解释性的基础上优化了分类规则.

Decision trees combination method based on genetic algorithm with application in CRM

For customer classification problems in customer relationship management (CRM), this paper proposes a combination classification method of multiple decision trees based on genetic algorithm. In the combination classification method, multiple decision trees that assign each class a probability measurement value are arranged with parallel combination. Genetic algorithm is used for the optimization of connection weight matrix in combination algorithm. Further more, the method is tested and evaluated by a simulation experiment that is conducted with simulation data of customer credit rating assessment. Compared with the single decision tree and other combination methods, experiment result shows that it improves the classification accuracy and optimizes classification rules, and holds favorable interpretability for classification result at the same time.